AI Stock Blowup and the Analyst Who Saw it Coming

Data is everywhere. When orchestrated properly, it sings. At Morningstar, we analyze and enrich data, making it actionable and powerful for you. Morningstar, where data speaks.

Supermicro doesn't really do the innovation. They are a contract manufacturer with willingness to commit working capital. Most of the innovation is done Upstream, Nvidia, TSMC.

Hello and welcome to the Barron Streetwise Podcast. I'm Jack Howe, and the voice you just heard, that's Mehdi Hosseini. He's an analyst at Susquehanna Financial Group, and he's talking about Super Microcomputers, which is not doing so super lately. There was a scathing report this past week from a well-known short seller, and the stock price has collapsed. And Mehdi was bearish all along. In a moment, we'll learn how he saw signs of trouble ahead of the crowd. Listening in is our audio producer Jackson. Hi Jackson. Hey Jack. You saw about the Hindenburg Report on Super Microcomputers this past week? Oh yeah. There's not a lot of companies out there that have taken Hindenburg as their brand name. That one hasn't really caught on, as I noted in this week's Barron's Magazine. Yeah, you don't want to name your company after a famous blimp disaster. Well, let's be careful with our terminology, please, Jackson. Blimps and zeppelins are two types of airships. I'm totally reading on the internet. And blimps are non-rigid, whereas zeppelins, that's what the Hindenburg was, was a rigid airship. There was a frame. But if it's more details that you're looking for on the subject— I wasn't. I can't provide you. With some of those. The Hindenburg disaster was 1937. It happened over Manchester Township, New Jersey. 7 million cubic feet of contained hydrogen ignited, and it brought down a German airship the size of a football field in a little over half a minute. And cameras caught the whole thing. One broadcaster said, this is the worst thing I've ever witnessed. And so you would not expect these days to find like Hindenburg brand peanut butter at the supermarket. Hindenburg Long Johns, just in time for winter. That's not gonna— nobody's looking for that. And certainly if you're a financial firm, you wouldn't want to have Hindenburg as your name. But if you're a short seller, if you're a short seller, it might be just the thing. Short sellers bet against stocks, and some of these professionals that have their own firms, they then publish reports, detailed reports, and they try to convince others that these stocks are unrelenting disasters. As I put it in this week's Barron's, a short seller would probably reject the name Titanic for his firm on the grounds that there were 706 survivors. So that's how we have a short-selling company called Hindenburg Research that published this report this past week on Super Microcomputer. And Hindenburg has a reputation. We've mentioned it on this podcast around 4 years ago. We had a conversation with Trevor Milton. Remember him, Jackson, the founder of that clean energy big rig startup Nikola? Yeah, I remember Nikola had been the subject at the time of a brutal report from Hindenburg, and Trevor Milton was pushing back on that. And he said that investors who bought his shares at the time would have, quote, one of the funnest rides they've ever had. And I wrote in Barron's that readers should sit this fund right out. And the shares since then are down 99%. And Milton late last year was sentenced to 4 years in prison for securities and wire fraud. Somewhat less dramatically, last year Hindenburg put out a report saying that the unit price of publicly traded Icahn Enterprises was, quote, inflated by 75%, and that one, since the report, is down 74%. So investors took notice this past week when Hindenburg put out a report saying about Supermicro that it has found, quote, glaring accounting red flags, including, quote, evidence of undisclosed related party transactions. We reached out to Supermicro. They said, quote, the company does not comment on rumors and speculation., but in a Wednesday filing, they said they would delay their annual report to assess, quote, the design and operating effectiveness of its internal controls over financial reporting. Those shares have collapsed from a high of over $1,200 in March to a recent $440. This episode will not be getting into any super sleuthing on the specific claims that Hindenburg has made. What I want to do instead is hear from someone who put out one of Wall Street's only bearish calls on the company so that maybe we as investors can learn more about how to spot these kinds of things early. I'm going to give you just a couple of quick points about what Supermicro does, and then we'll get to my conversation with Mehdi. Supermicro is what is known as an original design manufacturer, or ODM, and I think that's a category name that Sort of feels like it's overcompensating. What do those words suggest to you, Jackson? Original design manufacturer. Kind of sounds like a deli, like original beef sandwich. Exactly. If you're— if I see a place that says original beef sandwich, what I think is, okay, so there is some kind of dispute out there about who has the original beef sandwich. That tells me someone out there has another beef sandwich, at least one other, maybe two or three others, and there's some kind of argument here. So this one has to tell me that they're the original. It definitely does not tell me that this is a one-of-a-kind thing that I can't get anywhere else. Is that a fair way to put it? It sounds like you're the original gangster of deconstructing industry names. Thank you for that. So in this case, if you have an original design manufacturer, to me it just raises the question of how original is the thing that you do? Are you really like, are you starting with nothing? And then when you're done, there's something. And then presumably you could sell that something for high prices and collect fat profit margins. And that is definitely not what's going on here. Did I tell you, Jackson, about the time that I made a computer when I was in my 20s? Yeah. I still can't wrap my head around that. You don't seem like a computer-making guy.

Yeah.

That what I just said is true-ish, but kind of not totally because what I really did couldn't be called making a computer. It would be called more assembling a computer. In my 20s, my desire for computing oomph exceeded my budget oomph, and so I wanted to get a good computer on the cheap. And I looked at a website for cash-strapped nerds, and it listed all the components you need to just assemble your own computer. And I ordered the stuff, and I did it, and it worked. And the whole thing took about I'm gonna say 45 minutes and a screwdriver— the tool, not the drink. I'm not saying that I know as much or even nearly as much as Supermicro does about computing. I'm saying that Supermicro is sometimes lumped in with other AI stock plays like Nvidia, even though Nvidia knows a heck of a lot more than Supermicro about computing. And yes, you could say that that's opinion, but we could also use research and development spending as a proxy for smarts to get something other than opinion, something that we can measure. And if you look at Supermicro's latest financial results, which you would have to call preliminary results because they haven't yet put out their official full-year numbers yet, they show research and development spending of $348 million. I know that sounds like a lot of money, but it's about 2% of revenue. And NVIDIA spent $6 billion. That's 10% of a much larger revenue line. So if that's a measure of smarts, NVIDIA is a lot smarter. That's part of what you'll hear Mehdi explain in a moment. He'll also explain how Supermicro competes largely on the basis of its willingness to put up cash for the components to build its own brand of computing boxes that contain the high-value chips of its partners, like NVIDIA. But you'd much rather be NVIDIA, of course, than be the company who's putting up a lot of money to facilitate the sales of NVIDIA chips. What else? I'll mention two other things. One is about the financials, and I have to stress again that these are preliminary numbers. They are subject to being revised down the road. But if you look at the numbers for Supermicro's latest fiscal year, they show revenues roughly doubling to $14.9 billion, and they show earnings of $1.34 billion. That looks like a robust earnings line. You'd look at that and you'd say, Hey, this is a profitable company. There's good money being made here. But free cash flow for the same year was a negative $2.6 billion. And earnings and free cash flow, we've talked about this in the past. These are two measures that over time, they're measuring roughly the same thing. But in the short run, there could be big differences. Earnings pretend that revenues pair up neatly over each time period with the associated costs of those revenues. And that's to make it easy for stock investors to compare the two and see what's going on in terms of profitability. Free cash flow cares nothing about telling a neat story. It just measures how much money is coming in, how much money is going out. And investors tend to focus largely on earnings because they're easier to understand or make sense of. You can look at a chart and see a nice neat trajectory over time, but they would do well to also carefully look at free cash flow and compare the two. If you see a company that persistently has decent earnings and not so decent free cash flow, it's worth looking into. And this very high rate of cash burn, as you'll hear, comes from the fast growth in demand for artificial intelligence systems. If you partner with Nvidia to sell AI systems and you're putting up a lot of the cash for those systems and everybody wants to buy them, you need to put up a lot of money in a hurry. And that's a problem if you're operating in a business that doesn't have a ton of pricing power. Anyhow, the last thing I'll just say is that some AI-related companies, including NVIDIA, reported earnings this past week and the numbers look decent. They beat expectations, although people say they fell short of expectations about how much they would exceed expectations by— whisper numbers, if you like. And so the initial reaction for NVIDIA was the shares traded lower. You could say investors have got that wrong, the results are just fine. I'm sure we'll dig further into that in a future episode. But the point is, this is a moment of weakness for AI stocks that have been on a rip-roaring run. So you can imagine how that added to the negative reception for the Hindenburg report on Supermicro. Anyhow, let's take a quick break. When we come back, we'll hear my conversation with Mehdi Hosseini. Data is everywhere. When orchestrated properly, it sings. At Morningstar, we analyze and enrich data, making it actionable and powerful for you. Morningstar, where data speaks. Balance your trading strategy by adding futures. CME Group helps you manage risk and capture opportunities in all market environments. Capitalize on around-the-clock access to highly liquid global futures and options markets across all major asset classes. Visit your online broker to get started. See what adding futures can do for you at cmagroup.com/podcast. Derivatives are not suitable for all investors and involve risk of losing more than the amount originally deposited and any profit you may have made. This is not a recommendation or offer to buy, sell, or retain any specific investment or service. Welcome back. We're talking about Supermicro and artificial intelligence and NVIDIA and the Hindenburg short selling report. And we're going to hear now from Mehdi Hosseini. He's an analyst at Susquehanna Financial Group, and he's not just a Wall Street guy. He was in the biz, and by biz, I mean semiconductor design. I'll let Mehdi explain.

I have been covering tech for 24 years. Prior to that, I used to be a chip designer at National Semiconductor, which is now part of Texas Instruments. I have a master's degree in electrical engineering and an MBA.

Out of curiosity, how does the complexity of chip design compare with the complexity of accounting?

It's very— chip design is a lot more complicated and requires a lot of rigorous analytical work., and not sometimes all your work leads to favorable, uh, outcome. Wherein accounting is very analytical, and 9 out of 10 times you reach a concise outcome, is more black and white than designing circuit chips.

I'm glad that you said the chips are more complicated. That's good. So a lot of people watch the stock action with Supermicro and maybe aren't exactly sure of what the company does. Like, how would you sum it up, what they make and who they sell it to.

The company is better known for an outsourced contract manufacturer, but there has always been a differentiation. They have always benefited from new product introduction. Before AI and glorification of NVIDIA, Supermicro was a manufacturing partner for Intel and AMD. Every time Intel or AMD would have a new server CPU, Supermicro would be involved. They would see better than average growth in their revenue by building different SKUs or different variation of a server with the same CPU. So they would have this surge in revenue and earning and doing it for a little margin because they would try to mass customize a new CPU.

Why did it come to be that in the past couple of years the stock has rocketed higher, and we hear about this company in connection with AI. What change did they make, and how is the business different today?

Well, they took that strong partnership that they had with Intel and AMD and extended to NVIDIA, especially when NVIDIA was a scaling ampere, uh, in, in the first phase of AI adoption. Uh, Supermicro became successful, and, and they succeeded in becoming a partner with NVIDIA. Why wouldn't other Taiwan-based ODM be able to do the same thing, especially with NVIDIA that was trying to scale? Well, Supermicro, because they enable mass customization, they commit working capital. If they're building 10 servers based on the same CPU architecture, they put it on their balance sheet. Supermicro commits working capital. And the same thing happened with NVIDIA. And I think part of that partnership was with Supermicro, where to procure the key components. NVIDIA was scaling their Ampere and Supermicro became a reliable partner, especially committing working capital.

When you say committing working capital, but it means covered some of the expense upfront.

How would you describe that? You're exactly right. They will upfront the cost that associated with everything to do with a server box or a server rack. And that has to do with items outside of the core compute CPU or GPU.

That's attractive to a company like NVIDIA because it helps them manage their cash flow better. They don't have to put the money up. It allows them to take a sort of lighter approach in partnering. Okay, so I can see why that would be attractive. And that helped Supermicro quickly win business.

Yes, especially in the early days of AI and Ampere, which is 2 generations ago.

Now, is that a good way to do business, or is it a problem doing business like that long term? Or does it, does it depend? I mean, this approach of committing working capital, these kind of partnerships.

I don't think Supermicro changed their strategy that they have employed since their inception. I think it is investors that may have been thinking more optimistically, thinking that, oh, if Supermicro did 17% gross margin, and because Supermicro claims to be innovative, But this is sticky. And I think this is where the dislocation between what investors perceive, especially in the early days of AI, versus what the company's fundamentals are.

In other words, a company like NVIDIA can just go around to a different contract manufacturer and get a better deal. There's nothing that really keeps them with Supermicro. There's no proprietary advantage that Supermicro have.

Is that right?

Absolutely. I want to read to you the title of your own report. This is dated August 7th. Title of this report is Supermicro: Would It Take Another $2 Billion Plus of Free Cash Flow Burn to Double Revenue Again? So describe to me what that means, that the company is burning so much cash to create rapid revenue growth. How and why is it doing that? And why is that a problem?

Well, this cash burn, especially in the current environment, is a lot more elevated than in the prior cycle where the revenues were much lower. So as they scaled revenue, the cash burn has increased. Why so? Because they have to deal with the upfront costs. Again, there is no innovation. If you look at the R&D commitment, it's a fraction of what NVIDIA spends for R&D. It's a fraction of what TSMC commits for innovation. Look, my job is to be objective. When the company says, oh, we are innovating, no, you are raising capital because that's what's required for the upfront cost. So back to your question or comment about the title of that report, well, they burn a lot of cash to drive this much revenue. At the end of the day, it is the cash flow that determines the value of the equity or company, and they burn that much last fiscal year, and I think it's going to require that much to drive the same kind of growth this year. And I think this is something that investors haven't really grasped, perhaps until now.

Where's the money going to come from to keep burning cash like that?

They have raised equity a few times. They are borrowing. And I think if you take a look at the earning call transcript going back to early August, they did say, hey, We may have to raise more capital to support our growth. So the company is telling you, hey, we have to raise more capital to, to finance and support the revenue growth. It's just that the, the depth of the cash burn hasn't been understood until now.

You've looked at the stock for a long time. How would you characterize the trading in recent years? Some people would look at that chart and they'd say, oh, this must be a meme stock. The stock has gone absolutely crazy. Some people would say, no, it's just a, part of this AI, you know, people got very excited about AI and it's part of this basket of stocks. In the past, at some point you had a positive rating on the stock. You turned to a negative rating. What made you negative on the stock?

We as fundamental analysts, we're not really good in timing the inflection point in the stock. Our views are based on fundamental, very objective, and we constantly on a daily basis try to be very focused on objective analysis. So, um, I would, I would say that perhaps I was early, uh, with, with our negative view, but it was based on a fundamental view that this company requires a lot of cash and they have to burn a lot of cash. And it has taken perhaps 9 months for that thesis to play out. Um, now, who has been buying the stock, which back goes back to, uh, the first part of your question? I would say this is a meme stock. And I say that because I have gone through these hype cycles a few times with other sectors and other names. I used to have unpleasant emails and unpleasant voicemails in my phone from, I would say, investors that wouldn't be considered institutional investors.

Like a— This was not Warren Buffett leaving a message.

No, no, sir. I would say individual investors. That were, I wouldn't say they were pleasant and that's fine.

They were passionate. They were passionate people.

You couldn't characterize it any better. When you get this much passion involved with a stock without any analysis or fundamental hypothesis, then I would say, well, this is a meme stock. It's not based on analysis. It's based on passion and emotion.

—How do you tell? You could look at the stock and say, well, now it's half price versus a certain date in the past. You have a negative rating on the stock still. You have a sell rating. The stock as we speak is having another big down day at the time of this recording. How do you handle a situation like that? Because you work with cash flows and math and valuation and analysis, but how do you figure out when a stock like this is something that you're ready to turn more positive on Or do you just wipe your hands of it, you know, until there are, until there's some kind of transformation in the company management? In other words, is there a price that's low enough to be attractive or you're just not gonna go near it until something big changes in how the company's run?

That's actually a very good point. Yes, valuation matters. I've always argued this sector and this company specifically should at most be traded at one time. Enterprise value to sales, and we're still far from it. We still have a lot of downside risk to reach that valuation level in the best case. And on top of that, they just, as of early this morning, they're delaying the 10-K. And since this is not the first time, oh yeah, we need, we need some explaining. And you can't just say, oops, we're gonna fix it. There needs to be a deeper dive into how the company is structured.

Putting aside free cash flow, putting aside earnings, any of that, just the forecasts for revenue. There are people out there that anticipate fast revenue growth. Is that still in the cards for this company? When you have this kind of cycle of negativity, some customers might look at that and say, well, wait a second, we want to wait. Just as you say, we want to wait for more answers here before we know what to do. Is this the kind of situation where some of that anticipated revenue growth might now be at risk?

Absolutely. Look, why haven't the, the big blue chip hyperscalers like Azure, AWS, Google Cloud Service haven't become a big customer base or customer mix? And now you have the newcomers like CoreWeave. CoreWeave is a privately held AI service provider. They have options. NVIDIA has multiple options for a manufacturing partner. You're absolutely right. And this is why I think other than valuation, we need to wait for some clarity here because not only the way you recognize revenue and cost is in doubt, your customers are also going to doubt your viability. So you're absolutely correct by assuming that there needs to be more, uh, clarification. At the same time, customers may abandon this partner because of these irregularities. Unfortunately, I cannot quantify, but I would say, you know, you just asked me where is the valuation today? I said, in the best case scenario, one-time enterprise value to sales. Perhaps you want to give it a discount to account for these increased downside risk.

Moving beyond Supermicro, when you look across your coverage, you don't like this one. At the moment. Are there— can you name for us one or two stocks in your coverage that you are positive on, that you would call your favorites right now?

Yes, in the, in the hardware space, I like Pure Storage. What makes them stand out is they're actually doing a good job of penetrating hyperscalers.

The ticker there is PSTG. PSTG.

Yes. And they report tonight. So that's a positive rated stock. It's not really a direct peer to Supermicro, but it's in the same— they're downstream. They basically sell storage boxes, but much better financials and much better fundamentals. I do like companies that are the key innovator, like TSMC. Without TSMC, NVIDIA wouldn't even have an earnings conference call because they wouldn't be able to ship anything. TSMC as a key manufacturing partner of NVIDIA.

This is Taiwan Semiconductor, and for US investors, the ticker is TSM.

Yes, TSM. Taiwan Semiconductor Manufacturing. TSMC is the name, company, and TSM is the ticker. Look, there is a lot of, I would say, confusion on the risk, geopolitical risk. What if China bombs Taiwan? You know what? God forbid. If something happens and China were to bomb Taiwan, I think the global economy is just going to shut down. Without TSMC, Microsoft, Amazon's AWS, Google Cloud Services, they would just stop because they cannot get— they wouldn't be able to get the key components. Apple would not be able to sell iPhone. So I think the geopolitical risk is overhyped. TSMC or TSM is also a top-rated because they're they're the only company that could help NVIDIA and AMD and others to make these chips.

Thank you, Mehdi. I'm going to read a disclosure provided by Susquehanna Financial Group. SFG and/or its affiliates beneficially own 1% or more of the securities of Supermicro. That's something that has to do with the firm and not necessarily the analyst. Firms are required to put out those notices. Subscribe to the podcast on Apple Podcasts, Spotify, or wherever you listen to podcasts. If you listen on Apple, you can write us a review. You can email us with a question you'd like answered on the podcast. We've been getting loads of emails. Record on the voice memo app on your phone. You can send it to jack.hough@barons.com. I've been getting kind of a lot of email about the beard that I had for a little while and I shaved off. One person told me recently I look like Teen Wolf. Okay, that is ridiculous, sir. He had way better coverage in the high cheek area. Thanks and see you next week. Data is everywhere, but is it ready for consumption? Morningstar developed the language of global investment data so you have the right ingredients to help you shine. Morningstar, where data speaks.